Target Classification Method, Device, Electronic Device and Storage Medium

Abstract

The present invention discloses a target classification method, device, electronic device, and storage medium, relating to the field of image recognition technology. The method comprises: acquiring multiple target images to be processed; performing facial recognition on each target image to be processed to determine target images corresponding to the same person; and performing non-motor vehicle identification on each target image corresponding to each person to determine the number of different non-motor vehicles ridden by the same person. This method can determine the number of different non-motor vehicles included in the target image corresponding to the same person, and further determine the number of different non-motor vehicles ridden by the same person, thereby ensuring the accuracy of the number of different non-motor vehicles determined for the same person. Furthermore, the method improves the efficiency of identifying non-motor vehicles corresponding to the same person, saving time and labor costs.

Description

Technical Field

[0001] The present invention relates to the field of image recognition technology, and in particular to a target classification method, device, electronic device and storage medium. Background Art

[0002] With the surge in the number of non-motorized vehicles on the road, a new safety hazard has emerged. To ensure urban traffic safety, unified management of non-motorized vehicles is necessary. However, in current situations, the same person may be associated with multiple non-motorized vehicles. Therefore, it is necessary to identify multiple non-motorized vehicles associated with the same person.

[0003] However, in the prior art, relevant personnel are usually required to investigate each person's non-motor vehicle one by one and then record it.

[0004] Therefore, the above method is inefficient in recording multiple non-motor vehicles corresponding to the same person, wasting time and manpower costs. Summary of the Invention

[0005] In view of this, embodiments of the present invention provide a target classification method, device, electronic device, and storage medium, aiming to solve the problem of not being able to determine the number of non-motor vehicles corresponding to the same person.

[0006] According to the first aspect, an embodiment of the present invention provides a target classification method, which includes: obtaining multiple target images to be processed; performing face recognition on each target image to be processed to determine the target image corresponding to the same person; for the target image corresponding to each person, performing riding non-motor vehicle identification on each target image to determine the number of different non-motor vehicles ridden by the same person.

[0007] The target classification method provided by an embodiment of the present invention obtains multiple target images to be processed and performs facial recognition on each target image to determine the target image corresponding to the same person. This ensures the accuracy of the target images corresponding to each person and avoids errors in the number of different non-motor vehicles corresponding to the same person due to inaccurate target images corresponding to the same person. In addition, for each target image corresponding to the person, the non-motor vehicle identification performed on each target image can determine the number of different non-motor vehicles included in the target image corresponding to the same person, and further determine the number of different non-motor vehicles ridden by the same person, thus ensuring the accuracy of the number of different non-motor vehicles determined for the same person. Furthermore, relevant personnel are not required to conduct research and record the non-motor vehicles corresponding to each person. Therefore, the efficiency of identifying non-motor vehicles corresponding to the same person is improved, saving time and labor costs.

[0008] According to the second aspect, an embodiment of the present invention also provides a target classification device, which includes: an acquisition module for acquiring multiple target images to be processed; a first determination module for performing face recognition on each target image to be processed to determine the target image corresponding to the same person; a second determination module for performing riding non-motor vehicle identification on each target image corresponding to each person to determine the number of different non-motor vehicles ridden by the same person.

[0009] The target classification device provided by an embodiment of the present invention obtains multiple target images to be processed and performs facial recognition on each target image to determine the target image corresponding to the same person. This ensures the accuracy of the target images corresponding to each person and avoids errors in the number of different non-motor vehicles corresponding to the same person due to inaccurate target images corresponding to the same person. In addition, for each target image corresponding to the person, the non-motor vehicle ridden by each target image is identified, thereby determining the number of different non-motor vehicles included in the target image corresponding to the same person, and further determining the number of different non-motor vehicles ridden by the same person, thereby ensuring the accuracy of the number of different non-motor vehicles determined for the same person. Furthermore, there is no need for relevant personnel to conduct research and record the non-motor vehicles corresponding to each person. Therefore, the efficiency of identifying non-motor vehicles corresponding to the same person is improved, saving time and labor costs.

[0010] According to the third aspect, an embodiment of the present invention provides an electronic device, including a memory and a processor, the memory and the processor are communicatively connected to each other, computer instructions are stored in the memory, and the processor executes the target classification method in the first aspect or any one of the embodiments of the first aspect by executing the computer instructions.

[0011] According to a fourth aspect, an embodiment of the present invention provides a computer-readable storage medium, which stores computer instructions, and the computer instructions are used to enable a computer to execute the target classification method in the first aspect or any one embodiment of the first aspect. BRIEF DESCRIPTION OF THE DRAWINGS

[0012] In order to more clearly illustrate the specific embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the specific embodiments or the description of the prior art. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0013] Figure 1 This is a flow chart of a target classification method provided by an embodiment of the present invention;

[0014] Figure 2 is a flow chart of a target classification method provided by another embodiment of the present invention;

[0015] Figure 3 is a flow chart of a target classification method provided by another embodiment of the present invention;

[0016] Figure 4 This is a schematic diagram of face similarity comparison of candidate target images in a target classification method provided by another embodiment of the present invention;

[0017] Figure 5 is a flow chart of a target classification method provided by another embodiment of the present invention;

[0018] Figure 6 is a schematic diagram of non-motor vehicle attributes in a target classification method provided by another embodiment of the present invention;

[0019] Figure 7 is a flow chart of a target classification method provided by another embodiment of the present invention;

[0020] Figure 8 is a flow chart of a target classification method provided by another embodiment of the present invention;

[0021] Figure 9 This is a functional module diagram of a target classification device provided by an embodiment of the present invention;

[0022] Figure 10 It is a schematic diagram of the hardware structure of the electronic device provided by the embodiment of the present invention. DETAILED DESCRIPTION

[0023] To make the purpose, technical solutions, and advantages of the embodiments of the present invention more clear, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without making creative efforts shall fall within the scope of protection of the present invention.

[0024] It should be noted that the target classification method provided in the embodiment of the present application may be executed by a target classification device, which may be implemented as part or all of an electronic device through software, hardware, or a combination of software and hardware. The electronic device may be a server or a terminal. The server in the embodiment of the present application may be a single server or a server cluster composed of multiple servers. The terminal in the embodiment of the present application may be a smart phone, a personal computer, a tablet computer, a wearable device, an intelligent robot, or other intelligent hardware devices. In the following method embodiments, the execution subject is an electronic device as an example for explanation.

[0025] In one embodiment of the present application, Figure 1 As shown, a target classification method is provided, which is described by taking the application of the method to electronic equipment as an example, and includes the following steps:

[0026] S11. Acquire multiple target images to be processed.

[0027] In an optional implementation, the electronic device may acquire a plurality of target images to be processed sent by the image acquisition device based on the connection between the electronic device and the image acquisition device.

[0028] In another optional implementation, the electronic device may also utilize its own image acquisition device to acquire multiple target images to be processed.

[0029] In another optional implementation, the electronic device may receive a plurality of target images to be processed sent by other devices based on the connection between the electronic device and the other devices.

[0030] Exemplarily, the electronic device may obtain a plurality of target images to be processed captured by the traffic camera based on the connection with the traffic camera.

[0031] S12: Perform face recognition on each target image to be processed to determine the target image corresponding to the same person.

[0032] Specifically, the electronic device can use the target detection model to perform face recognition on each target image to be processed, and then compare the facial feature information corresponding to each identified target image to be processed with the facial feature information corresponding to each person in the database, so as to determine the target image corresponding to the same person.

[0033] Among them, the target detection model can be a model based on manual features, such as DPM (Deformable Parts Model), and the target detection model can also be a model based on convolutional neural networks, such as YOLO (You Only Look Once) detector, R-CNN, (Region-based Convolutional Neural Networks, region-based convolutional neural network) model, SSD (Single Shot MultiBox, single shot multi-box) detector and Mask R-CNN (Mask Region-based Convolutional Neural Networks, masked region-based convolutional neural network) model, etc. The embodiment of the present application does not specifically limit the target detection model.

[0034] For example, the electronic device can receive a target image to be processed from a traffic camera, then identify the face in the target image to determine facial features. The identified facial features are then compared with corresponding facial features in a database to determine that the face in the current target image is person A. Using the above method, facial recognition is performed on multiple target images to determine the person corresponding to the face in each target image. Target images corresponding to the same person are then grouped together to obtain target images corresponding to the same person.

[0035] S13. For each target image corresponding to each person, identify the target image as a person riding a non-motorized vehicle, and determine the number of different non-motorized vehicles ridden by the same person.

[0036] Specifically, for each target image corresponding to each person, the electronic device can identify each target image as a non-motor vehicle being ridden, and then determine the number of different non-motor vehicles ridden by the same person based on the feature information of the non-motor vehicle corresponding to each target image.

[0037] For a detailed description of this step, please refer to the following implementation method.

[0038] The target classification method provided by an embodiment of the present invention obtains multiple target images to be processed and performs facial recognition on each target image to determine the target image corresponding to the same person. This ensures the accuracy of the target images corresponding to each person and avoids errors in the number of different non-motor vehicles corresponding to the same person due to inaccurate target images corresponding to the same person. In addition, for each target image corresponding to the person, the non-motor vehicle identification performed on each target image can determine the number of different non-motor vehicles included in the target image corresponding to the same person, and further determine the number of different non-motor vehicles ridden by the same person, thus ensuring the accuracy of the number of different non-motor vehicles determined for the same person. Furthermore, relevant personnel are not required to conduct research and record the non-motor vehicles corresponding to each person. Therefore, the efficiency of identifying non-motor vehicles corresponding to the same person is improved, saving time and labor costs.

[0039] In an optional embodiment of the present application, Figure 2 As shown, a target classification method is provided, which is described by taking the application of the method to electronic equipment as an example, and includes the following steps:

[0040] S21, obtaining multiple target images to be processed. S21 may include the following steps:

[0041] S211. Acquire multiple images to be processed.

[0042] In an optional implementation, the electronic device may acquire a plurality of to-be-processed images sent by the image acquisition device based on the connection between the electronic device and the image acquisition device.

[0043] In another optional implementation, the electronic device may also utilize its own image acquisition device to acquire multiple images to be processed.

[0044] In another optional implementation, the electronic device may receive a plurality of to-be-processed images sent by other devices based on the connection between the electronic device and other devices.

[0045] Exemplarily, the electronic device may obtain a plurality of images to be processed captured by the traffic camera based on the connection with the traffic camera.

[0046] S212: Perform target recognition on each image to be processed to determine the type of non-motor vehicle in the image to be processed and the distance between the non-motor vehicle and the person.

[0047] Specifically, the electronic device can use the target detection model to perform target recognition on each target image to be processed, where target recognition can include human and non-motor vehicle recognition. Based on the target recognition results, the electronic device can determine the type of non-motor vehicle and the distance between the non-motor vehicle and the person.

[0048] In the embodiment of the present invention, the target detection model can be a model based on manual features, such as DPM (Deformable Parts Model), and the target detection model can also be a model based on convolutional neural network, such as YOLO (You Only Look Once) detector, R-CNN (Region-based Convolutional Neural Networks, region-based convolutional neural network) model, SSD (Single Shot MultiBox, single shot multi-box) detector and Mask R-CNN (Mask Region-based Convolutional Neural Networks, masked region-based convolutional neural network) model, etc. The embodiment of the present application does not specifically limit the target detection model.

[0049] S213 , screening the multiple images to be processed according to the type of the non-motorized vehicle and the distance between the non-motorized vehicle and the person to obtain multiple target images to be processed.

[0050] Specifically, the electronic device can screen multiple images to be processed according to the non-motorized vehicle type and the distance between the non-motorized vehicle and the person, eliminate the images to be processed corresponding to the specific non-motorized vehicle type and the images to be processed whose non-motorized vehicle type and the distance between the non-motorized vehicle and the person do not meet the requirements, and obtain the target image to be processed.

[0051] For example, the electronic device may eliminate the images to be processed whose types of non-motor vehicles are shared bicycles, tricycles, children's vehicles, etc. according to the types of non-motor vehicles corresponding to the images to be processed.

[0052] The electronic device can also eliminate images where the distance between the bottom of the non-motor vehicle and the bottom of the person is less than 0.06 based on the distance between the non-motor vehicle and the person in each image to be processed. This is because when the bottom of the person is closely aligned with the non-motor vehicle, the non-motor vehicle is often incomplete, resulting in inaccurate feature values and low accuracy of the comparison results. The non-motor vehicle area and the person area do not overlap, making it difficult to identify the non-motor vehicle corresponding to the person in the image to be processed.

[0053] In an optional embodiment of the present application, the electronic device may further recognize a face in the image to be processed and, based on the recognized facial feature information, determine whether the age of the person corresponding to the face in the image to be processed is less than 12 years old. If the age of the person corresponding to the face in the image to be processed is less than 12 years old, the image to be processed is discarded.

[0054] In an optional embodiment of the present application, since a small facial pixel area can easily lead to facial recognition errors, to improve facial recognition accuracy, the electronic device can further calculate the facial pixel area in the image to be processed. If the facial pixel area is smaller than a preset area, the image to be processed is discarded. The preset area can be 5000 or 6000, and this embodiment of the application does not specifically limit the preset area.

[0055] In an optional embodiment of the present application, since the recognition of non-motor vehicles will be affected when non-motor vehicles are close to the edge of the image to be processed, the electronic device can identify the non-motor vehicles in the image to be processed, determine the position of the non-motor vehicles in the image to be processed, and eliminate the images to be processed in which non-motor vehicles are close to the edge of the image to be processed.

[0056] S22: Perform face recognition on each target image to be processed to determine the target image corresponding to the same person.

[0057] For details about this step, please refer to the above Figure 1 S12 in the description will not be described here.

[0058] S23. For each target image corresponding to each person, identify the target image as a person riding a non-motorized vehicle, and determine the number of different non-motorized vehicles ridden by the same person.

[0059] For details about this step, please refer to the above Figure 1 S22 in the description will not be described here.

[0060] The target classification method provided by an embodiment of the present invention acquires multiple images to be processed and performs target recognition on each image to determine the type of non-motorized vehicle and the distance between the non-motorized vehicle and a person in the image to be processed. The multiple images to be processed are then filtered based on the non-motorized vehicle type and the distance between the non-motorized vehicle and the person. This eliminates images that are not of the target non-motorized vehicle type and images that do not meet the required distance between the non-motorized vehicle and the person, thereby ensuring the accuracy of the acquired target images to be processed.

[0061] In an optional embodiment of the present application, Figure 3 As shown, a target classification method is provided, which is described by taking the application of the method to electronic equipment as an example, and includes the following steps:

[0062] S31. Acquire multiple target images to be processed.

[0063] For details about this step, please refer to the above Figure 2 S21 in the article will not be described in detail here.

[0064] S32: Perform face recognition on each target image to be processed to determine the target image corresponding to the same person.

[0065] In an optional implementation, the above step S32 may include the following steps:

[0066] S321 , performing face recognition on each target image to be processed to obtain a face recognition result.

[0067] Specifically, the electronic device can use the target detection model to perform face recognition on each target image to be processed to obtain a face recognition result.

[0068] Among them, the target detection model can be a model based on manual features, such as DPM (Deformable Parts Model), and the target detection model can also be a model based on convolutional neural networks, such as YOLO (You Only Look Once) detector, R-CNN, (Region-based Convolutional Neural Networks, region-based convolutional neural network) model, SSD (Single Shot MultiBox, single shot multi-box) detector and Mask R-CNN (Mask Region-based Convolutional Neural Networks, masked region-based convolutional neural network) model, etc. The embodiment of the present application does not specifically limit the target detection model.

[0069] S322. Determine the labeling information of the person corresponding to the target image to be processed based on the face recognition result.

[0070] Specifically, the electronic device can determine the facial feature information corresponding to the target image to be processed based on the face recognition result, and then compare the facial feature information corresponding to each identified target image to be processed with the facial feature information corresponding to each person in the database to determine the labeling information of the person corresponding to the target image to be processed.

[0071] S323: Determine the target image corresponding to the same person according to the person's labeled information.

[0072] In an optional implementation, the electronic device determines the target images to be processed with the same person labeling information as target images corresponding to the same person.

[0073] In an optional implementation, the above step S323 may further include the following steps:

[0074] (1) Determine candidate target images based on the person’s annotation information.

[0075] (2) Calculate the face similarity of any two candidate target images.

[0076] (3) If the similarities of all faces are greater than or equal to the first preset similarity threshold, the candidate target image is retained.

[0077] (4) If there is at least one face whose similarity is less than the first preset similarity threshold, the candidate target image is deleted.

[0078] (5) Determine the target image corresponding to the same person based on the retained candidate target images.

[0079] Specifically, after the labeling information of the person corresponding to each target image to be processed is determined, candidate target images corresponding to the same person can be determined based on the person labeling information.

[0080] In an optional embodiment, the electronic device may perform facial recognition on each candidate target image separately, determine facial feature information corresponding to each candidate target image, and then calculate similarity based on the facial feature information corresponding to any two candidate target images.

[0081] In another optional implementation, the electronic device may perform face recognition on each candidate target image separately, and then calculate the face similarity between any two candidate target images using a Euclidean distance calculation method.

[0082] If the facial similarity of each face in the candidate target image is greater than or equal to the first preset similarity threshold, the candidate target image is retained; if at least one face has a similarity less than the first preset similarity threshold, the candidate target image is deleted. The electronic device can determine the target image corresponding to the same person based on the retained candidate target images. The first preset similarity threshold can be 0.7 or 0.6, and this application does not specifically limit the first preset similarity threshold.

[0083] For example, Figure 4As shown. Suppose there are 5 candidate target images, and the first preset similarity threshold is 0.7. The electronic device recognizes the 5 candidate target images respectively, and determines the facial feature values corresponding to the 5 candidate target images. Then the facial feature values corresponding to the 5 candidate target images are compared in pairs. Among them, the facial similarity between the facial feature value A corresponding to the first candidate target image and the facial feature value B corresponding to the second candidate target image is 0.83; the facial similarity between the facial feature value A corresponding to the first candidate target image and the facial feature value C corresponding to the third candidate target image is 0.92; the facial similarity between the facial feature value A corresponding to the first candidate target image and the facial feature value D corresponding to the fourth candidate target image is 0.69, and the remaining comparison results are no longer listed one by one. It can be seen that if there is a face whose similarity is less than the first preset similarity threshold, the electronic device deletes these 5 candidate target images.

[0084] In an optional embodiment of the present application, the above-mentioned “calculating the face similarity of any two candidate target images” may further include the following steps:

[0085] (21) Obtain the quality level of each candidate target image.

[0086] (22) According to the quality level, determine the candidate target images whose quality level is greater than a preset quality threshold.

[0087] (23) Calculate the face similarity of any two candidate target images whose quality levels are greater than a preset quality threshold.

[0088] Specifically, to ensure the accuracy of the target image corresponding to the same person and improve the efficiency of identifying each candidate target image, the electronic device can also detect the quality of each candidate target image to obtain a quality level of each candidate target image. The quality of each candidate target image can be related to clarity, contrast, facial position, non-motor vehicle position, etc. of each candidate target image.

[0089] After obtaining the quality level of each candidate target image, the electronic device can determine, based on the quality level, candidate target images whose quality level exceeds a preset quality threshold. To improve the efficiency of identifying each candidate target image, the electronic device can only calculate facial similarity for any two candidate target images whose quality level exceeds the preset quality threshold. The preset quality threshold can be 4 or 5, and the embodiments of this application do not specifically limit the preset quality threshold.

[0090] If the facial similarities corresponding to any two candidate target images with a quality level greater than the preset quality threshold are both greater than or equal to the first preset similarity threshold, the candidate target images with a quality level greater than the preset quality threshold are retained; if there is at least one facial similarity less than the first preset similarity threshold, it is determined that the candidate target image currently determined based on the person's annotation information may be abnormal, that is, the face corresponding to the current candidate target image may not be the same person. In order to ensure the accuracy of the number of different non-motor vehicles corresponding to the same person finally determined, the candidate target image needs to be deleted.

[0091] S33. For each target image corresponding to each person, identify the target image as a person riding a non-motorized vehicle, and determine the number of different non-motorized vehicles ridden by the same person.

[0092] For details about this step, please refer to the above Figure 2 S23 in the description will not be described here.

[0093] The target classification method provided by the embodiment of the present invention performs face recognition on the target image to obtain a face recognition result, and then determines the labeling information of the person corresponding to the target image based on the face recognition result, and then determines the target image corresponding to the same person based on the labeling information of the person, so that the target image corresponding to the same person can be determined quickly and accurately, thereby improving the accuracy of classifying the image to be processed based on the person labeling information.

[0094] Then, the target classification method provided by an embodiment of the present invention determines candidate target images based on the person's annotated information and calculates facial similarity between any two candidate target images. This allows the electronic device to determine whether to retain or discard a candidate target image based on the facial similarity between the two candidate target images. If the facial similarity between any two candidate target images is greater than or equal to a first preset similarity threshold, the faces in each candidate target image can be accurately determined to be the same person, and the candidate target images are retained. Based on the retained candidate target images, target images corresponding to the same person are determined. This ensures the accuracy of the number of different non-motor vehicles corresponding to the same person determined based on the target images. If at least one facial similarity is less than the first preset similarity threshold, it can be determined that the candidate target images currently determined based on the person's annotated information may be anomaly, meaning that the face corresponding to the current candidate target image may not be the same person. To ensure the accuracy of the final determination of the number of different non-motor vehicles corresponding to the same person, the candidate target image needs to be deleted. Therefore, the above target classification method ensures the accuracy of the target images determined to be the same person, and further ensures the accuracy of the determination of the number of different non-motor vehicles corresponding to the same person.

[0095] In addition, the target classification method provided by the embodiment of the present invention obtains the quality level of each candidate target image, so that the candidate target images with a quality level greater than a preset quality threshold can be determined based on the quality level. The facial similarity of any two candidate target images with a quality level greater than the preset quality threshold is calculated, and then whether to retain the candidate target image is determined based on the facial similarity corresponding to each candidate target image, thereby ensuring the accuracy of the target images corresponding to the same person. In addition, if the facial similarity of any two candidate target images with a quality level greater than the preset quality threshold is less than the first preset similarity threshold, then the facial similarity of any two candidate target images with a quality level less than the preset quality threshold does not need to be calculated and can be determined to be less than the first preset similarity threshold, thereby reducing the processing process of the candidate target images and improving the efficiency of target classification.

[0096] In an optional embodiment of the present application, Figure 5 As shown, a target classification method is provided, which is described by taking the application of the method to electronic equipment as an example, and includes the following steps:

[0097] S41. Acquire multiple target images to be processed.

[0098] For details about this step, please refer to the above Figure 3 S31 in the figure is not described here.

[0099] S42: Perform face recognition on each target image to be processed to determine the target image corresponding to the same person.

[0100] For details about this step, please refer to the above Figure 3 S32 in the figure is not described here.

[0101] S43. For each target image corresponding to each person, identify the target image as a person riding a non-motorized vehicle, and determine the number of different non-motorized vehicles ridden by the same person.

[0102] In an optional embodiment of the present application, the above S43 may include the following steps:

[0103] S431 , performing non-motor vehicle recognition on each target image to obtain a non-motor vehicle recognition result.

[0104] S432. Calculate the similarity of non-motor vehicles corresponding to any two target images based on the non-motor vehicle recognition result.

[0105] S433: If the similarity between the non-motor vehicles corresponding to any two target images is less than a second preset similarity threshold, determine that the two target images correspond to different non-motor vehicles.

[0106] S434. Determine the number of different non-motor vehicles ridden by the same person based on the comparison result.

[0107] Specifically, the electronic device can use the target detection model to perform non-motor vehicle recognition on each target image to obtain a non-motor vehicle recognition result. Then, based on the non-motor vehicle recognition result, the electronic device can determine the non-motor vehicle feature information corresponding to each target image, and then calculate the similarity between the non-motor vehicle feature information corresponding to any two target images.

[0108] In another optional implementation, the electronic device may perform non-motor vehicle identification on each target image separately, and then use the Euclidean distance calculation method to calculate the similarity between any two target images.

[0109] If the similarity between the non-motor vehicles corresponding to any two target images is less than a second preset similarity threshold, the two target images are determined to correspond to different non-motor vehicles. Based on the comparison results, the number of different non-motor vehicles ridden by the same person is determined. The second preset similarity threshold can be 0.3 or 0.35, and this embodiment of the application does not impose a specific limitation on the second preset similarity threshold. The second preset similarity threshold can be obtained through statistical calculation based on the purity of the algorithm.

[0110] For example, assuming there are three target images and the second preset similarity threshold is 0.3, the electronic device may compare the non-motor vehicle feature information in the first target image with the non-motor vehicle feature information in the second target image to determine the similarity between the non-motor vehicles corresponding to the first target image and the second target image. If the similarity between the non-motor vehicles corresponding to the first target image and the second target image is less than 0.3, the first target image and the second target image are determined to correspond to different non-motor vehicles. The electronic device may then compare the non-motor vehicle feature information in the first target image with the non-motor vehicle feature information in the third target image to determine the similarity between the non-motor vehicles corresponding to the first target image and the third target image. If the similarity between the non-motor vehicles corresponding to the first target image and the third target image is also less than 0.3, the first target image and the second target image are determined to correspond to different non-motor vehicles. The electronic device can then compare the non-motor vehicle characteristic information in the second target image with the non-motor vehicle characteristic information in the third target image to determine the similarity between the non-motor vehicles corresponding to the second and third target images. If the similarity between the non-motor vehicles corresponding to the second and third target images is also less than 0.3, the second and third target images are determined to be different non-motor vehicles. Thus, the electronic device determines that the number of different non-motor vehicles ridden by the same person is three.

[0111] In an optional implementation of the present application, if the similarity of non-motor vehicles corresponding to any two target images is greater than or equal to the second preset similarity threshold and less than the fourth similarity threshold, the electronic device compares the attributes of the non-motor vehicles in the target images whose non-motor vehicle similarity is less than the fourth similarity threshold and greater than the second preset similarity threshold. If the attributes of the non-motor vehicles are not exactly the same, the electronic device determines that the target images corresponding to the target images correspond to different non-motor vehicles. Based on the comparison results, the number of different non-motor vehicles ridden by the same person is determined. Among them, the attribute information of the non-motor vehicle can be as follows: Figure 6 shown.

[0112] In an optional implementation of the present application, the above S434 may include the following steps:

[0113] (1) According to the comparison results, the target images corresponding to different non-motor vehicles are selected;

[0114] (2) performing face recognition on the screened target image, and determining whether a similarity threshold between facial feature information corresponding to the screened target image is greater than a third similarity threshold;

[0115] (3) If yes, then the number of different non-motor vehicles ridden by the same person is determined based on the number of target images corresponding to different non-motor vehicles.

[0116] Specifically, in order to improve the accuracy of determining the number of different non-motor vehicles ridden by the same person, the electronic device may further select target images corresponding to different non-motor vehicles based on the comparison results.

[0117] The electronic device can then use the target detection model to perform facial recognition on the selected target images and determine whether a similarity threshold between the facial feature information corresponding to the selected target images is greater than a third similarity threshold. It should be noted that the third similarity threshold is greater than the first similarity threshold. The third similarity threshold can be 0.8 or 0.85.

[0118] If the similarity threshold between the facial feature information corresponding to the screened target images is greater than the third similarity threshold, the number of different non-motor vehicles ridden by the same person is determined based on the number of target images corresponding to different non-motor vehicles.

[0119] For example, assume that there are four target images and the third similarity threshold can be 0.8. After performing non-motor vehicle recognition on the target images as described above and calculating the similarity of the non-motor vehicles corresponding to any two target images, the comparison result is that the 1st to 3rd images of the four target images correspond to different non-motor vehicles, and the non-motor vehicle corresponding to the 4th target image is the same as the non-motor vehicle corresponding to the 1st target image. Therefore, the electronic device selects three target images corresponding to different non-motor vehicles from the 4th target image. Then, face recognition is performed on the three target images corresponding to different non-motor vehicles. If the similarity of the faces corresponding to these three target images is greater than 0.8, the electronic device determines that the same person has ridden three different non-motor vehicles.

[0120] The target classification method provided by an embodiment of the present invention performs non-motor vehicle identification on each target image to obtain a non-motor vehicle identification result. Then, based on the non-motor vehicle identification result, the similarity between any two target images is calculated. This ensures that the number of different non-motor vehicles included in the target image can be determined based on the non-motor vehicle similarity between the target images. If the similarity between any two target images is less than a second preset similarity threshold, the two target images are determined to correspond to different non-motor vehicles. Based on the comparison results, the number of different non-motor vehicles ridden by the same person is determined. This ensures the accuracy of the number of different non-motor vehicles determined for the same person.

[0121] In addition, the target classification method provided by the embodiment of the present invention screens target images corresponding to different non-motor vehicles based on the comparison results, then performs facial recognition on the screened target images, and determines whether the similarity threshold between the facial feature information corresponding to the screened target images is greater than a third similarity threshold. If so, the number of different non-motor vehicles ridden by the same person is determined based on the number of target images corresponding to different non-motor vehicles. The above-mentioned target classification method, since facial recognition is performed again on the target images corresponding to different non-motor vehicles, and it is determined whether the similarity threshold between the facial feature information corresponding to the screened target images is greater than the third similarity threshold, can ensure that the target images corresponding to different non-motor vehicles correspond to the same person, further ensuring the accuracy of the number of different non-motor vehicles corresponding to the same person.

[0122] In an optional embodiment of the present application, the target classification method may further include the following steps:

[0123] S71. Calculate the similarity of non-motor vehicles corresponding to any two target images based on the non-motor vehicle recognition result.

[0124] Specifically, the electronic device can use the target detection model to perform non-motor vehicle recognition on each target image to obtain a non-motor vehicle recognition result. Then, based on the non-motor vehicle recognition result, the electronic device can determine the non-motor vehicle feature information corresponding to each target image, and then calculate the similarity between the non-motor vehicle feature information corresponding to any two target images.

[0125] In another optional implementation, the electronic device may perform non-motor vehicle identification on each target image separately, and then use the Euclidean distance calculation method to calculate the similarity between any two target images.

[0126] S72: If the similarity between the non-motor vehicles corresponding to any two target images is greater than a fourth similarity threshold, it is determined that the two target images correspond to the same non-motor vehicle.

[0127] Specifically, if the similarity between any two target images of non-motor vehicles is greater than a fourth similarity threshold, the two target images are determined to correspond to the same non-motor vehicle. The fourth similarity threshold can be 0.5 or 0.55, and is not specifically limited in this embodiment of the present application. The fourth similarity threshold can be obtained through statistical calculation based on the purity of the algorithm.

[0128] S73 : If the similarity between the non-motor vehicles corresponding to any two target images is less than or equal to the fourth similarity threshold and greater than the fifth similarity threshold, obtain attribute information of the non-motor vehicles corresponding to the two target images.

[0129] The fourth similarity threshold is greater than the fifth similarity threshold.

[0130] Specifically, if the similarity between any two target images of non-motor vehicles is less than or equal to the fourth similarity threshold and greater than the fifth similarity threshold, the attribute information of the non-motor vehicles corresponding to the two target images is retrieved. The fifth preset similarity threshold can be 0.3 or 0.35, and is not specifically limited in this embodiment of the present application. The fifth preset similarity threshold can be obtained through statistical calculation based on the purity of the algorithm.

[0131] S74. Determine, based on the attribute information, that the two target images correspond to the same non-motor vehicle.

[0132] In an optional embodiment of the present application, the above step S74 may further include the following steps:

[0133] S741: Compare the attribute information of the non-motor vehicle corresponding to the two target images.

[0134] S742: If the attribute information of the non-motor vehicles corresponding to the two target images is the same, it is determined that the two target images correspond to the same non-motor vehicle;

[0135] S743: If the attribute information of the non-motor vehicles corresponding to the two target images is different, determine that the two target images correspond to different non-motor vehicles.

[0136] Specifically, the electronic device compares the attribute information of the non-motor vehicles corresponding to the two target images. If the attribute information of the non-motor vehicles corresponding to the two target images is the same, it is determined that the two target images correspond to the same non-motor vehicle; if the attribute information of the non-motor vehicles corresponding to the two target images is different, the electronic device determines that the two target images correspond to different non-motor vehicles.

[0137] S75. According to the determination result, the same non-motor vehicles ridden by the same person are classified into one category.

[0138] Specifically, the electronic device classifies the same non-motor vehicles ridden by the same person into one category based on the above determination result.

[0139] The target classification method provided by an embodiment of the present invention compares the similarity of non-motor vehicles corresponding to any two target images. When the similarity of non-motor vehicles corresponding to any two target images is greater than a fourth similarity threshold, it is determined that the two target images correspond to the same non-motor vehicle. This allows the same non-motor vehicle corresponding to the same person to be classified. When the similarity of non-motor vehicles corresponding to any two target images is less than or equal to the fourth similarity threshold and greater than a fifth similarity threshold, the attribute information of the non-motor vehicles corresponding to the two target images is compared, allowing the electronic device to determine that the two target images correspond to the same non-motor vehicle based on the attribute information of the non-motor vehicle. This further enables the same non-motor vehicles corresponding to the same person to be classified into one category, ensuring the accuracy of the classification of non-motor vehicles corresponding to the same person.

[0140] Furthermore, the target classification method provided by an embodiment of the present invention compares the attribute information of non-motor vehicles corresponding to two target images. If the attribute information of the non-motor vehicles corresponding to the two target images is the same, the two target images are determined to correspond to the same non-motor vehicle; if the attribute information of the non-motor vehicles corresponding to the two target images is different, the two target images are determined to correspond to different non-motor vehicles. This ensures the accuracy of non-motor vehicle classification corresponding to the same person and avoids the situation where two non-motor vehicles with different attribute information are classified into the same category.

[0141] In another optional embodiment of the present application, Figure 7As shown, after determining the number of different non-motor vehicles corresponding to the same person, the electronic device can retain the target images whose non-motor vehicle similarity is greater than the second preset similarity threshold, and then perform face recognition on the retained target images, and again only retain the target images whose face similarity threshold is greater than the third similarity threshold to obtain the target images to be classified.

[0142] In another optional embodiment of the present application, Figure 8 It is shown that the electronic device can archive the riding event for the target image to be classified. The electronic device can perform quality identification on the target image to be classified, and determine the high-quality target image to be classified and the low-quality target image to be classified corresponding to the person. Then, the target image to be classified whose quality is greater than the preset threshold is determined as a high-quality image, and the target image to be classified whose quality is less than or equal to the preset threshold is determined as a low-quality image. The preset threshold can be 0.5 or 0.4. The embodiment of the present application does not specifically limit the preset threshold.

[0143] The electronic device identifies non-motor vehicles on high-quality target images to be classified, and classifies target images to be classified whose similarity to non-motor vehicles is greater than a fourth similarity threshold into one category. The fourth similarity threshold can be 0.5 or 0.46, and the present embodiment does not specifically limit the fourth similarity threshold.

[0144] The electronic device compares the attributes of the non-motor vehicles in the target images whose similarity to the non-motor vehicles is less than the fourth similarity threshold and greater than the second preset similarity threshold. If the attributes of the non-motor vehicles are exactly the same, the target images with exactly the same non-motor vehicle attributes are classified into one category.

[0145] For example, assuming the fourth similarity threshold is 0.5 and the second preset similarity threshold is 0.3, the electronic device performs non-motor vehicle identification on high-quality target images to be classified and classifies target images with a non-motor vehicle similarity greater than 0.5 into one category. The electronic device then continues to compare the non-motor vehicle attributes corresponding to target images with a non-motor vehicle similarity greater than 0.3 but less than 0.5. If the non-motor vehicle attributes are identical, the target images with identical non-motor vehicle attributes are classified into one category.

[0146] During the archiving process, each file type can select a target image with the highest quality as the file cover. Unarchived target images are not displayed in the file.

[0147] According to the above-mentioned principle of archiving high-quality target images, low-quality target images are archived. In this way, not only the number of non-motor vehicles corresponding to the same person can be determined, but also the non-motor vehicles corresponding to the same person can be archived.

[0148] It should be understood that although Figure 1-3 、 Figure 5 as well as Figure 7-8 The steps in the flowchart are shown in sequence as indicated by the arrows, but these steps are not necessarily executed in the order indicated by the arrows. Unless otherwise specified in this document, there is no strict order restriction for the execution of these steps, and these steps can be executed in other orders. In addition, Figure 1-3 、 Figure 5 as well as Figure 7-8 At least part of the steps may include multiple steps or multiple stages. These steps or stages are not necessarily performed at the same time, but can be performed at different times. The order of execution of these steps or stages is not necessarily one by one, but can be performed in turn or alternately with other steps or at least part of the steps or stages in other steps.

[0149] like Figure 9 As shown, this embodiment provides a target classification device, which includes:

[0150] An acquisition module 51 is used to acquire a plurality of target images to be processed;

[0151] The first determination module 52 is used to perform face recognition on each target image to be processed to determine the target image corresponding to the same person;

[0152] The second determining module 53 is configured to identify the target images corresponding to the respective persons as those riding non-motor vehicles, and determine the number of different non-motor vehicles ridden by the same person.

[0153] In one embodiment of the present application, the acquisition module 51 is specifically used to acquire multiple images to be processed; perform target recognition on each image to be processed, determine the type of non-motor vehicle in the image to be processed and the distance between the non-motor vehicle and the person; filter the multiple images to be processed according to the type of non-motor vehicle and the distance between the non-motor vehicle and the person, and obtain multiple target images to be processed.

[0154] In one embodiment of the present application, the above-mentioned first determination module 52 is specifically used to perform face recognition on each target image to be processed to obtain a face recognition result; determine the labeling information of the person corresponding to the target image to be processed based on the face recognition result; and determine the target image corresponding to the same person based on the labeling information of the person.

[0155] In one embodiment of the present application, the above-mentioned first determination module 52 is specifically used to determine candidate target images based on the person's annotation information; calculate the facial similarity of any two candidate target images; if the facial similarities of each face are greater than or equal to the first preset similarity threshold, retain the candidate target image; if there is at least one face similarity less than the first preset similarity threshold, delete the candidate target image; and determine the target image corresponding to the same person based on the retained candidate target images.

[0156] In one embodiment of the present application, the above-mentioned first determination module 52 is specifically used to obtain the quality level of each candidate target image; determine the candidate target images whose quality level is greater than a preset quality threshold based on the quality level; and calculate the face similarity of any two candidate target images whose quality level is greater than the preset quality threshold.

[0157] In one embodiment of the present application, the above-mentioned second determination module 53 is specifically used to perform non-motor vehicle identification on each target image to obtain a non-motor vehicle identification result; based on the non-motor vehicle identification result, calculate the similarity of the non-motor vehicles corresponding to any two target images; if the similarity of the non-motor vehicles corresponding to any two target images is less than a second preset similarity threshold, it is determined that the two target images correspond to different non-motor vehicles; based on the comparison result, determine the number of different non-motor vehicles ridden by the same person.

[0158] In one embodiment of the present application, the above-mentioned second determination module 53 is specifically used to screen target images corresponding to different non-motor vehicles based on the comparison results; perform face recognition on the screened target images, and determine whether the similarity threshold between the facial feature information corresponding to the screened target images is greater than the third similarity threshold; if so, determine the number of different non-motor vehicles ridden by the same person based on the number of target images corresponding to different non-motor vehicles.

[0159] For the specific limitations and beneficial effects of the target classification device, please refer to the limitations of the target classification method above and will not be repeated here. The various modules in the above-mentioned target classification device can be implemented in whole or in part by software, hardware, or a combination thereof. The above-mentioned modules can be embedded in or independent of the processor in the electronic device in the form of hardware, or can be stored in the memory of the electronic device in the form of software, so that the processor can call and execute the operations corresponding to the above modules.

[0160] An embodiment of the present invention further provides an electronic device having the above Figure 9 The target classification device shown.

[0161] like Figure 10 As shown, Figure 10 is a structural diagram of an electronic device provided by an optional embodiment of the present invention, such as Figure 10As shown, the electronic device may include: at least one processor 61, such as a CPU (Central Processing Unit), at least one communication interface 63, a memory 64, and at least one communication bus 62. The communication bus 62 is used to realize the connection and communication between these components. The communication interface 63 may include a display screen (Display), a keyboard (Keyboard), and the optional communication interface 63 may also include a standard wired interface and a wireless interface. The memory 64 may be a high-speed RAM memory (Random Access Memory, volatile random access memory) or a non-volatile memory (non-volatile memory), such as at least one disk memory. The memory 64 may optionally be at least one storage device located away from the aforementioned processor 61. The processor 61 may be combined with Figure 9 In the described apparatus, the memory 64 stores an application program, and the processor 61 calls the program code stored in the memory 64 to execute any of the above method steps.

[0162] The communication bus 62 may be a peripheral component interconnect (PCI) bus or an extended industry standard architecture (EISA) bus. The communication bus 62 may be divided into an address bus, a data bus, a control bus, etc. For ease of representation, Figure 6 Only one thick line is used in the diagram, but this does not mean that there is only one bus or one type of bus.

[0163] Among them, the memory 64 may include volatile memory (English: volatile memory), such as random-access memory (English: random-access memory, abbreviated: RAM); the memory may also include non-volatile memory (English: non-volatile memory), such as flash memory (English: flash memory), hard disk drive (English: hard disk drive, abbreviated: HDD) or solid-state drive (English: solid-state drive, abbreviated: SSD); the memory 64 may also include a combination of the above types of memory.

[0164] The processor 61 may be a central processing unit (CPU), a network processor (NP), or a combination of a CPU and a NP.

[0165] The processor 61 may further include a hardware chip. The hardware chip may be an application-specific integrated circuit (ASIC), a programmable logic device (PLD), or a combination thereof. The PLD may be a complex programmable logic device (CPLD), a field-programmable gate array (FPGA), a generic array logic (GAL), or any combination thereof.

[0166] Optionally, the memory 64 is also used to store program instructions. The processor 61 can call the program instructions to implement the application Figures 1 to 3 as well as Figure 5 The target classification method shown in the embodiment.

[0167] An embodiment of the present invention further provides a non-transitory computer storage medium storing computer-executable instructions that can execute the target classification method in any of the above method embodiments. The storage medium can be a magnetic disk, an optical disk, a read-only memory (ROM), a random access memory (RAM), a flash memory, a hard disk drive (HDD), or a solid-state drive (SSD); the storage medium can also include a combination of the above types of memory.

[0168] Although the embodiments of the present invention have been described with reference to the accompanying drawings, those skilled in the art may make various modifications and variations without departing from the spirit and scope of the present invention. Such modifications and variations are all within the scope defined by the appended claims.

Claims

1. A target classification method, characterized in that: The method comprises: Acquire multiple target images to be processed; Performing face recognition on each of the target images to be processed to obtain a face recognition result; Determining the labeling information of the person corresponding to the target image to be processed according to the face recognition result; Determining a candidate target image based on the labeled information of the person; Calculating the facial similarity of any two candidate target images; If the similarities of the faces are all greater than or equal to a first preset similarity threshold, retaining the candidate target image; If there is at least one face whose similarity is less than the first preset similarity threshold, deleting the candidate target image; Determining a target image corresponding to the same person based on the retained candidate target images; For each target image corresponding to each person, performing non-motor vehicle recognition on each target image to obtain a non-motor vehicle recognition result; Calculating the similarity of non-motor vehicles corresponding to any two target images based on the non-motor vehicle recognition result; If the similarity of the non-motor vehicles corresponding to any two target images is less than a second preset similarity threshold, it is determined that the two target images correspond to different non-motor vehicles; screening the target images corresponding to different non-motor vehicles according to the comparison results; Performing face recognition on the screened target image to determine whether a similarity threshold between facial feature information corresponding to the screened target image is greater than a third similarity threshold; If yes, determining the number of different non-motor vehicles ridden by the same person based on the number of target images corresponding to different non-motor vehicles; The step of obtaining a plurality of target images to be processed includes: Acquire multiple images to be processed; Performing target recognition on each of the images to be processed to determine the type of non-motor vehicles in the images to be processed and the distance between the non-motor vehicles and the person; screening the plurality of images to be processed according to the type of the non-motor vehicle and the distance between the non-motor vehicle and the person to obtain a plurality of target images to be processed; The method further comprises: The non-motor vehicle in the image to be processed is identified, the position of the non-motor vehicle in the image to be processed is determined, and the image to be processed in which the non-motor vehicle is at the edge of the image to be processed is eliminated.

2. The method according to claim 1, characterized in that The calculating of face similarity between any two candidate target images includes: Obtaining a quality level of each candidate target image; Determining, according to the quality level, the candidate target images whose quality level is greater than a preset quality threshold; The face similarity is calculated for any two candidate target images whose quality levels are greater than a preset quality threshold.

3. The method according to claim 1, characterized in that The method further comprises: Calculating the similarity of non-motor vehicles corresponding to any two target images based on the non-motor vehicle recognition result; If the similarity of the non-motor vehicles corresponding to any two target images is greater than a fourth similarity threshold, it is determined that the two target images correspond to the same non-motor vehicle; If the similarity between the non-motor vehicles corresponding to any two of the target images is less than or equal to the fourth similarity threshold and greater than a fifth similarity threshold, then acquiring the attribute information of the non-motor vehicles corresponding to the two target images; wherein the fourth similarity threshold is greater than the fifth similarity threshold; determining, based on the attribute information, that the two target images correspond to the same non-motor vehicle; Based on the determination results, the same non-motor vehicles ridden by the same person are classified into one category.

4. The method according to claim 3, characterized in that Determining that the two target images correspond to the same non-motor vehicle according to the attribute information includes: Comparing the attribute information of the non-motor vehicle corresponding to the two target images, If the attribute information of the non-motor vehicles corresponding to the two target images is the same, determining that the two target images correspond to the same non-motor vehicle; If the attribute information of the non-motor vehicles corresponding to the two target images is different, it is determined that the two target images correspond to different non-motor vehicles.

5. A target classification device, characterized in that: The device comprises: An acquisition module is configured to acquire a plurality of target images to be processed; wherein acquiring the plurality of target images to be processed comprises: acquiring a plurality of images to be processed; performing target recognition on each of the images to be processed, determining the type of non-motor vehicles in the images to be processed and the distance between the non-motor vehicles and the personnel; screening the plurality of images to be processed according to the type of non-motor vehicles and the distance between the non-motor vehicles and the personnel to obtain a plurality of target images to be processed; wherein the method further comprises: identifying the non-motor vehicles in the images to be processed, determining the positions of the non-motor vehicles in the images to be processed, and eliminating the images to be processed in which the non-motor vehicles are at the edges of the images to be processed; A first determination module is configured to perform face recognition on each of the target images to be processed to obtain a face recognition result; determine, based on the face recognition result, the labeling information of the person corresponding to the target image to be processed; determine a candidate target image based on the labeling information of the person; calculate face similarity between any two of the candidate target images; retain the candidate target images if the face similarities of each of the candidate target images are greater than or equal to a first preset similarity threshold; delete the candidate target images if at least one face similarity is less than the first preset similarity threshold; and determine, based on the retained candidate target images, a target image corresponding to the same person; The second determination module is used to perform non-motor vehicle identification on the target image corresponding to each person to obtain a non-motor vehicle identification result; calculate the similarity of the non-motor vehicles corresponding to any two target images based on the non-motor vehicle identification result; if the similarity of the non-motor vehicles corresponding to any two target images is less than a second preset similarity threshold, determine that the two target images correspond to different non-motor vehicles; based on the comparison result, filter the target images corresponding to different non-motor vehicles; perform face recognition on the filtered target images to determine whether the similarity threshold between the facial feature information corresponding to the filtered target images is greater than a third similarity threshold; if so, determine the number of different non-motor vehicles ridden by the same person based on the number of target images corresponding to different non-motor vehicles.

6. An electronic device, characterized in that: The device comprises a memory and a processor, wherein the memory stores computer instructions, and the processor executes the target classification method according to any one of claims 1 to 4 by executing the computer instructions.

7. A computer-readable storage medium, characterized in that The computer-readable storage medium stores computer instructions, and the computer instructions are used to enable a computer to execute the target classification method according to any one of claims 1 to 4.

Images